Originally Posted by
Tomaski
The Seattle Times article made this important point:
When the first reports of this testing came out there was a tremendous amount of discussion and speculation about the FCC processors, how they got overwhelmed and the "fact" that the electric trim switches were slow to respond. Appears that discussion was all based on some reporters misunderstanding, so before the current discussion goes to far afield it might be worth remembering this. Even the best news outlets may not have the best information and it is very hard to understand deep technical details of this system without direct access to the experts and their data. Certainly frustrating as our understanding continues to evolve but a good caution against becoming to wedded to any particular theory.
Unfortunately the information blobs still don't line up. So the fault scenario seems to be a bit flip. I'm not sure about the number five. I dont't really believe they tested for five flips coinciding. Maybe a total of five scenarios involving a bit flip somewhere. Anyway, the AND trim output persits and does not seem to be stoppable w/o cutting off. This seems odd. One would expect that either the FCC power latches after detecting a memory parity error - taking a safe output state during power latch - or continues operating, overwriting the flipped bit with new information when the state changes, I.e. manual electric trim is applied. Neither obviously does happen so this could very well be the point where the "overwhelmed", "sluggish" part somehow comes into play.
The Seattle Times articles by Mr. Gates are quality wise way ahead of the products of other news outlets. However after this one I am not having less questions then before.